Andrey Babushkin focused on stabilizing the test environment for the openvinotoolkit/nncf repository by resolving Python dependency conflicts in quantization_aware_training_torch tests. He addressed environment-related test failures by conditionally installing dependencies such as whowhatbench only when required, and enhanced diagnostics through improved logging around Python requirements installation. Using his skills in Python scripting, dependency management, and testing, Andrey improved test reproducibility and observability, which reduced flaky CI runs and enabled faster debugging. His work contributed to more deterministic test outcomes and higher confidence in release readiness, demonstrating a methodical approach to Python packaging and test instrumentation challenges.